Re: [PATCH v9 2/3] scsi: ufs: core: Hold host_sem in ufshcd_wl_poweroff()

[Peter Wang]

On Tue, 2026-11-10 at 11:03 +0900, HOYOUNG SEO wrote:
> ufshcd_wl_suspend() takes hba->host_sem and ufshcd_wl_resume()
> releases
> it, which serializes the error handler against system suspend and
> resume. ufshcd_wl_poweroff(), the hibernation poweroff callback, also
> suspends the device with UFS_SHUTDOWN_PM but does not take host_sem.
> An
> error handler that runs in the meantime can reset the link
> concurrently
> with __ufshcd_wl_suspend().
> 
> It also leaves host_sem unbalanced: if hibernation is aborted after
> ufshcd_wl_poweroff() has been called, ufshcd_wl_resume() is called as
> the .restore callback and releases host_sem, which was not acquired.
> 
> Make ufshcd_wl_poweroff() behave like ufshcd_wl_suspend() and keep
> holding host_sem. It is released by ufshcd_wl_resume() if hibernation
> is
> aborted. If __ufshcd_wl_suspend() fails, release host_sem and return
> the
> error as ufshcd_wl_suspend() does, otherwise the error handler, which
> waits for host_sem, would block forever.
> 
> This changes the behavior on failure: the error is now returned from
> the
> .poweroff callback, which aborts hibernation, instead of being
> ignored.
> 
> Do not use hba->shutting_down here: it makes
> ufs_mtk_system_suspend(),
> which is also the .poweroff callback of the MediaTek host controller,
> fail with -EBUSY, and nothing clears it when hibernation is aborted.
> 
> Fixes: b294ff3e3449 ("scsi: ufs: core: Enable power management for
> wlun")
> Suggested-by: Bart Van Assche <bvanassche@xxxxxxx>
> Signed-off-by: HOYOUNG SEO <hy50.seo@xxxxxxxxxxx>
> ---

Reviewed-by: Peter Wang <peter.wang@xxxxxxxxxxxx>